Vibrothermography, also known as thermosonics or sonic infrared, is a method of nondestructive evaluation that finds cracks or delaminations from the heat given off in response to vibration. In vibrothermography, finding cracks requires identifying and localizing pulsed surface and subsurface heat sources from a time sequence of infrared images. Traditionally this identification involves manually stepping through and studying the images. Careful observation of the heating and subsequent cooling is needed to distinguish cracks from false indications. In this paper, we present an algorithm that reduces the entire time sequence to a single static plot. The plot uses only a few coefficients per pixel to reconstruct the original sequence; this is possible because the reduction is based on a physical model. As an added bonus, the algorithm reduces noise and improves sensitivity. A single false-color image summarizes all the information from the entire sequence, simplifying the task of identifying cracks.